RetargetingKit maps live or recorded human motion to robot hands, arms, and humanoids using RoboKit optimization and neural hand models.

The two main applications are:

• 🎮 Real-time teleoperation
• ⚡ Offline retargeting in parallel

Real-time teleoperation with dexterous hands and humanoids.

Body-and-hand retargeting in a furnished scene.

Quick links

• Homepage
• Docs
• Quickstart
• Installation
• Development

Installation

Requires Python 3.10 or newer:

uv add retargetingkit      # or: pip install retargetingkit

See Installation for source installs, extras, and asset access.

Features

Hand retargeting	Geometry-aware hand retargeting	Real-time hand teleoperation
Humanoid retargeting	Geometry-aware humanoid retargeting	Real-time humanoid teleoperation
Full-body retargeting	Geometry-aware full-body retargeting	Real-time full-body teleoperation

Explore recorded motions and robot presets in the gallery examples.

How it works

flowchart LR
    Source["`Source
    *VR · gloves · mocap · recordings*`"]
    Retargeter["`Retargeter
    *Optimization + Neural*`"]
    Target["`Target
    *Viser · robot hardware*`"]
    Source -- "Frame: hand · arm · body" --> Retargeter
    Retargeter -- "Command: robot joints · base pose" --> Target
    classDef endpoint fill:#dbeafe,stroke:#2563eb,color:#111827
    classDef solver fill:#dbc8ea,stroke:#8163a1,stroke-width:2px,color:#111827
    class Source,Target endpoint
    class Retargeter solver

Loading

Component	Role	Read more
Source	Read a device or recording and produce a canonical frame.	Sources
Frame	Represent human hand, arm, or body motion.	Frames
Retargeter	Solve hand, arm, or body motion using RoboKit.	Interfaces
Command	Carry robot joint positions and an optional base pose.	Commands
Target	Preview commands or send them to a robot.	Targets

See the framework overview for frames, commands, and named rigs.